The number of knee surgeries is expected to rise significantly due to aging and obesity. Load-bearing 2D X-rays, the gold standard for knee surgery planning, suffer from projection and magnification errors. The advent of weightbearing CT provides better anatomical accuracy by capturing joint alignment under loading conditions in three dimensions. However, like conventional CT, they are costly, time-consuming, and expose patients to higher radiation than X-rays. Hence, there remains a clear clinical need to combine the strengths of both X-ray and CT, improving patient assessment while maintaining accessibility and efficiency. This PhD research develops AI-driven solutions to enhance patient selection and surgical planning. The first objective focuses on automating anatomical landmark detection on X-rays and CT scans using deep learning, improving measurement accuracy and clinical insights. The second objective aims to extract 3D anatomical information solely from 2D X-rays using advanced generative AI models. By integrating expertise from orthopedic clinicians and AI researchers, the project ensures its methods align with real-world obstacles in surgical decision-making and practical challenges in patient care. Leveraging large, multicenter datasets, the research will undergo rigorous clinical validation.